Influence of spatial heterogeneity and temporal variability in habitat selection: A case study on a great bustard metapopulation

We modelled great bustard abundance patterns and their spatial structure in relation to habitat and landscape variables. We developed Generalized Linear Models (GLM) using long term data series – years 1997–2006 – during the breeding season in Madrid region, central Spain. Our main goal was to assess spatial and temporal variability effects on habitat selection in this species evaluating the impact of interannual variability on habitat selection, the consistency of model predictions among years and the effect of data accumulation in the model performance. We examined the predictive ability of our models using internal and external validation techniques. We built separate models for each year and different models using the addition of several year data. The model perfomance increased as more census data were included in the calibration. One-off temporal data was insufficient to predict great bustard abundance properly. The final model (calibrated with all period data) showed a reasonable accuracy, attending to the validation tests. The variability in habitat suitability predictions between annual models does not seem to be caused by changes in habitat selection between years because the global model had a better explanatory ability than annual models. As far as interannual variability in spring greenness is concerned, the most variable sites are preferred, suggesting a selection for sites with smaller land use units and with a traditional rotation system. The great bustard abundance in Madrid was affected by the presence of other conspecifics but this pattern was conditioned to the existence of a suitable habitat denoted by the other variables in the final model. Future persistence of great bustards in Madrid region depends on a sustainable economic development that maintains traditional land uses, at least in areas with high ecological value for great bustards, whether they are occupied or not.

► We developed annual models that predict great bustard abundance in spring.
► We assessed spatial, monthly and interannual variability effects on habitat selection.
► One-off temporal data was insufficient to predict great bustard abundance properly.
► Sites with higher stability in vegetation biomass in one spring were selected.
► The most variable sites in the interannual spring grennes are preferred.